Roll LLVM b20d7d02..6811a83c815#6844
Conversation
|
There are some LLVM IR changes that are hard for me to judge but do not look trivial. For instance in |
jonmeow
left a comment
jonmeow
left a comment
There was a problem hiding this comment.
Approving to give the chance to see if this helps with the latest sha issue.
| // CHECK:STDOUT: | ||
| // CHECK:STDOUT: !0 = !{i32 7, !"Dwarf Version", i32 5} | ||
| // CHECK:STDOUT: !1 = !{i32 2, !"Debug Info Version", i32 3} | ||
| // CHECK:STDOUT: !2 = !{i32 1, !"wchar_size", i32 4} |
There was a problem hiding this comment.
FYI things are churning because wchar_size is being removed. That's not too important, it's not something we're explicitly adding, just something llvm was dropping in.
github-merge-queue
bot
removed this pull request from the merge queue due to failed status checks
|
FYI in #6854 I'm trying to switch to git_override to work around the sha256 issues. |
By using git_override, we get some validation from the sha, while removing the sha256 on the .tar.gz which has been brittle lately. Note the difference between downloading via sha is this still locally validates content. Versus something like #6844, this doesn't update the llvm version, just how we get it. Assisted-by: Google Antigravity with Gemini
|
Gah, sorry i missed that this had passed checks before hitting merge on the |
| // Let diagnostics fall out of scope and be destroyed. This finishes | ||
| // diagnosing any failures before we verbosely log the source of those | ||
| // failures. |
There was a problem hiding this comment.
Optional (happy for it here, or in a follow-up, or I can send a follow-up):
Given that upstream is now baking more assumptions in about the scope here, I would be inclined to extract this block into a helper function? That would let it return the integer and failing commands, and also hold a comment about the fact that the compilation is fundamentally stateful. WDYT?
3 participants
Roll LLVM to
6811a83c81500ee373adfc0d9978ff9625a4cf1c.This includes llvm/llvm-project#183831 which moved the functionality of
finish()onDiagnosticConsumers into the destructors, and removed thefinish()method. So, our callers tofinish()are migrated to cause the destructor to run at that time instead.